One-step synthesis of highly pure and well-crystallized vertically aligned carbon nanotubes

HAL Id: cea-03002425

https://hal-cea.archives-ouvertes.fr/cea-03002425

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

One-step synthesis of highly pure and well-crystallized

vertically aligned carbon nanotubes

Emeline Charon, Mathieu Pinault, Martine Mayne-L’hermite, Cécile Reynaud

To cite this version:

Emeline Charon, Mathieu Pinault, Martine Mayne-L’hermite, Cécile Reynaud. One-step synthesis of highly pure and well-crystallized vertically aligned carbon nanotubes. Carbon, 2020, 173, pp.758-768. �10.1016/j.carbon.2020.10.056�. �cea-03002425�

One-step synthesis of highly pure and well-crystallized vertically aligned

1

carbon nanotubes

2

Emeline Charon*, Mathieu Pinault, Martine Mayne-L’Hermite, Cécile Reynaud 3

Université Paris-Saclay, CEA, CNRS, NIMBE, 91191 Gif sur Yvette Cedex, France 4

5

Abstract (197 words)

6

The one-step aerosol-assisted catalytic chemical vapor deposition (CCVD) process, when 7

operated in a H2-based carrier gas, is shown to be effective at an extremely low ferrocene

8

content (0.1wt.% in toluene), which enables to efficiently prepare forests of vertically aligned 9

multiwalled carbon nanotubes (VACNTs) with high purity and good crystalline level. The 10

resulting iron content in VACNT sample is only 0.5wt.% corresponding to a high catalytic yield 11

of 200 while the steady growth rate (20 µm/min) is maintained at a significant value. The iron 12

content in the sample is found proportional to the ferrocene concentration in the precursor, and 13

when it decreases, a noticeable reduction in the iron encapsulation frequency in the nanotube 14

channel is plainly observed. Lowering the ferrocene concentration generates a reduction in the 15

VACNT number density while the growth rate increases and CNT diameters and crystalline 16

structure remain similar. Such significant purity and crystalline levels involves a high oxidation 17

resistance of the VACNT forest. These results, together with the in-situ continuous growth 18

feature of the one-step aerosol assisted CCVD process, open up towards a new VACNT range 19

made of pure and well-crystallized multi-walled carbon nanotubes, which is of interest for 20

industrial production and commercial applications. 21

22

Keywords: vertically aligned carbon nanotubes; one-step catalytic chemical vapour deposition;

23

iron content; purity; crystallinity; oxidation resistance. 24

1 Introduction

26

Vertically Aligned Carbon NanoTubes (VACNT) arrays have demonstrated outstanding 27

potentialities as a novel material to prepare nanostructured composite or multifunctional 28

materials. The carbon nanotube (CNT) alignment allows to take advantage of their anisotropic 29

properties. This is useful for many applications of high societal interest such as heat transfer [1], 30

energy storage [2,3], or high performance structural composites [4,5]. Efforts are currently 31

underway to develop efficient synthesis processes for the manufacture of VACNTs in order to 32

achieve mass production while reducing costs [6,7]. In particular, the process must be as simple 33

as possible and must be operated at atmospheric pressure. The most effective method for 34

synthesizing VACNT is catalytic chemical vapor deposition (CCVD) [8] which can be operated 35

at atmospheric pressure. However, it often involved several preliminary steps at low pressure 36

to prepare the layer of catalytic nanoparticles from which the nanotubes will then grow. It is 37

why the aerosol-assisted version of the CCVD process has been developed [9–14], where 38

catalyst nanoparticles are produced in situ and continuously refreshed due to the simultaneous 39

injection of both metal and carbon precursors[15]. This process involves a base-growth 40

mechanism [16] enabling the continuous growth of VACNTs at steady growth rate over very 41

long synthesis duration, which allows VACNTs greater than 1mm to be achieved. Therefore, 42

aerosol-assisted CCVD is a versatile, cost-effective and scalable one-step process operated at 43

atmospheric pressure potentially applicable for a roll to roll production of multi-walled 44

VACNTs. The continuous supply of ferrocene, which is essential to sustain the growth through 45

the refreshment of the catalyst nanoparticles [16], induces the encapsulation of short nanowires 46

in the inner channel of the nanotubes [17,18]. As a consequence, the iron content of the pristine 47

VACNTs (Table 1) is typically around a few wt.% [15,19–21]. Such metallic by-products can 48

alter the intrinsic nanotube properties and be detrimental for several applications [22]. In 49

addition, it is also recognized that they can catalyze the oxidation of nanotubes and thus affect 50

their stability in air at moderate temperatures [23–25]. Therefore, these CNTs need to be 51

purified and several methods have been developed to remove metal by-products [22], either by 52

acid treatment [26], but their complete removal is difficult to achieve without altering the 53

nanotube structure [27,28], or by high temperature annealing which is the most effective 54

method for their removal while improving the nanotube crystallinity [22,29,30]. However, such 55

treatments complicate the CNT production process and are not always compatible with the 56

targeted application. 57

Therefore, efforts must be done directly on the aerosol-assisted CCVD production process in 58

order to reduce drastically the iron content in the pristine VACNTs. Recently and in the 59

meantime of our study, this issue was addressed by T. Kinoshita et al. [31] who reported a 60

modified floating catalyst CCVD method (similar to the aerosol-assisted one) enabling the 61

growth of CNT forests containing very low concentration of impurity (0.8 wt%). However, the 62

changes made to the process have turned it into a two-step process involving in the first step 63

the in-situ formation of the catalyst particles, then in the second step the CNT growth with 64

carbon precursors only. Such experimental conditions do not allow the refreshment of the 65

catalyst particles so that the lifetime of their process is limited. Moreover the second step was 66

performed at low pressure which limits the industrial interest of the method. 67

In this context, the aim of our work is to reduce the Fe content of pristine VACNTs while 68

preserving their crystallinity, alignment and growth rate by increasing the catalytic yield of the 69

one-step aerosol-assisted CCVD process operated at atmospheric pressure. In the literature 70

(Table 1), few studies [15,19–21] report the measurement of iron content of VACNTs formed 71

by the one-step aerosol-assisted CCVD, but it is generally found that iron content increases 72

with ferrocene concentration in the precursor solution [19,21]. Cho et al. [20] studied the [0.2-73

9] wt.% ferrocene range at 760°C and obtained 1.9wt.% as the lowest iron content in CNT 74

samples for 1.5wt.% of ferrocene in toluene. However, they did not observe any growth of 75

VACNT forests at 0.2wt.% of ferrocene, in agreement with the result of Singh et al. [13]. 76

Similarly, Castro et al. [15] observed no growth of VACNT at 0.1 and 0.5wt.% when studying 77

the [0.1-5] wt.% range of ferrocene at 850°C and the lowest iron content obtained is 3wt.% at 78

2.5wt% ferrocene with a higher growth rate due to a higher synthesis temperature. To our 79

knowledge, for one-step aerosol-assisted CCVD process, the lowest concentration of ferrocene 80

responsible for the growth of VACNTs is 0.1wt.% [32] but the growth rate at 750°C is 81

particularly low (0.4 µm/min) and the iron content in the resulting sample is not reported. The 82

highest growth rate obtained with a low ferrocene concentration (0.58wt.%) is 13.6 µm/min at 83

780°C [33] but the iron content is not indicated. The latter result was obtained with 15% H2

84

added to Ar as carrier gas. 85

86 87 88 89

90 91

Table 1 : Lowest ferrocene concentrations giving rise to VACNT growth reported in the

92

literature for the one-step aerosol-assisted CCVD synthesis. The most relevant

93

experimental parameters are reported, as well as, the growth rate and iron content of the

94

final product when available. Ferrocene concentrations in the precursor solution were

95

converted to wt.%. Note that 1wt.% of ferrocene in toluene corresponds to a Fe/C ratio

96 of 0.07at.%. 97 Ferrocene concentration (wt.%) Carbon source Carrier gas T (°C) Substrate Growth rate (µm/min) Iron content (wt.%) Raman ID/IG  (nm) Bai 2005 [33] 0.58 xylene Ar/H2 87/13 780 quartz 13.6 - - 1.14 20.4 - Ionescu 2011 [32] 0.1 xylene Ar 750 Sputtered Al Si wafer 0.4 - 0.49 785 nm 1 4.7 0.68 785 nm Cho 2012 [20] 1.5 toluene Ar 760 quartz 2.5 1.9 0.33* 633 nm McKee 2009 [19] 0.54 benzene Ar 750 various - 0.7 - 3 0.3 measured at 1.75 wt% Fe 514 nm Meysami 2013 [21] 1 toluene Ar 800 quartz 16 2.5 0.29 532 nm Castro 2010 [15] 1.25 toluene Ar 850 quartz 11 3.5 0.52* 532 nm 2.5 20 3 0.33* 532 nm Castro 2013 [34] 2.5 toluene Ar/H2 80/20 850 quartz 29 5.3 - 800 23 6.2 - * Area ratio 98

Despite the wide disparity in the experimental conditions, the results of the literature highlight 99

the important role played by the synthesis temperature and the hydrogen presence in the carrier 100

gas to maintain a realistic growth rate while decreasing the ferrocene concentration in the 101

aerosol. In a previous paper [34], we studied the effect of adding hydrogen to Ar carrier gas and 102

showed that it has a major impact on the aerosol-assisted CCVD process. Increasing the 103

hydrogen content strongly shifts the CNT growth to the reactor inlet and reduces the CNT 104

diameters, while the CNT number density increases. Indeed, hydrogen demonstrably affects the 105

gas phase phenomena by lowering the decomposition temperature of ferrocene, promoting the 106

formation of a larger number of smaller iron nanoparticles in the gas phase and increasing the 107

nucleation site density of the catalytic nanoparticles. In addition, the optimal synthesis 108

temperature in terms of yield and homogeneity of growth is reduced to 800°C under hydrogen 109

(Table 1) [34]. In these previous experiments, the concentration of ferrocene in toluene was set 110

at 2.5wt.%, the optimal concentration determined at 850°C without H2 [15]. It should be

111

reminded that without H2, no growth of VACNTs is obtained below 1.25wt.% of ferrocene,

112

neither at 850°C nor at 800°C. However, the situation may change in the presence of H2 due to

113

the significant improvement in the ferrocene decomposition it induces. Therefore, our objective 114

is to determine, in the presence of H2, the lowest concentration of ferrocene in toluene allowing

115

the growth of a VACNT forest with low impurity level while maintaining a high growth rate 116

and a good structural quality. 117

In this paper, we investigate the role of ferrocene concentration in the precursor solution 118

(toluene/ferrocene) on the VACNT growth and characteristics (iron content, length, diameter, 119

density, structure, oxidation resistance as well as on growth efficiency (yields and growth rate). 120

As hydrogen was reported to be a determining chemical player in the decomposition rate of 121

ferrocene [34], all the synthesis of this study are performed by the one-step aerosol assisted 122

CCVD process with hydrogen diluted in argon as carrier gas. The approach combines global 123

analysis by Scanning Electron Microscopy (SEM), Thermogravimetric Analysis (TGA) and 124

Raman Spectrometry, with local analysis by Transmission Electron Microscopy (TEM). Since 125

in this one-step aerosol assisted CCVD process a continuous supply of ferrocene is essential to 126

sustain growth [35], our first objective is to determine the ferrocene concentration threshold 127

above which VACNT growth is maintained once the catalytic particles are formed. To this end, 128

we first carry out double injection experiments, by designing a specific set-up allowing to use 129

a solution containing 2.5wt.% ferrocene in a first short injection sequence to form the catalyst 130

particles, and then to use a second solution with a lower ferrocene concentration in a second 131

long injection sequence. These experiments demonstrate that a very low ferrocene 132

concentration (0.01wt.%) used in the second sequence can sustain the growth of VACNT forest, 133

which is considerably lower than the ones reported in the literature for this one-step aerosol 134

assisted CCVD process. Subsequently, by operating this process in a single injection mode with 135

a ferrocene concentration decreasing from 2.5 to 0.01wt.%, we show that it is possible to 136

determine conditions to synthesize a pristine VACNT carpet with unprecedented reported low 137

iron content (0.5 wt%) and high oxidation resistance. 138

2 Experimental

139

2.1 Synthesis

140

The set-up (Fig. 1) and standard procedure used to synthesize VACNTs by the one-step aerosol-141

assisted CCVD process at atmospheric pressure are similar to those described by Castro et 142

al. [34]. Note that syntheses in the presence of H2 cannot be done under the same conditions as

143

in its absence because the addition of H2 considerably shifts the VACNT growth towards the

144

reactor inlet. The synthesis temperature must then be lowered to restore growth in the reactor 145

center. This is why all syntheses here are made in the presence of H2 and at the optimal

146

temperature of 800°C [34]. The furnace exhibits an isothermal area larger than the previous 147

one. It is a 45 cm long tubular furnace equipped with a quartz reactor (16 mm inner diameter). 148

At 800°C, the isothermal area (±10°C), is extended from 10 to 35 cm from the furnace inlet 149

(Fig. 1). A flat quartz substrate of 1.5x0.75 cm2 is placed in the isothermal area at 14 cm from 150

the reactor inlet. The VACNTs grown on this substrate are analyzed together with the VACNTs 151

grown on the reactor walls. The VACNTs grown on the reactor walls are collected with a 152

stainless scraper, a procedure which keeps intact their morphology and structure. They are 153

weighed according to the location in the reactor, divided into 5 cm long sections with a surface 154

area of 25 cm2 each, in order to assess the distribution of the VACNT deposit. All the 155

characterizations will be carried out on the VACNTs coming from the substrate of zone 2 and 156

the VACNT powder coming from zone 3. These zones were chosen because the yield is better 157

there. To perform double injection experiments, the present set-up is equipped with two liquid 158

precursor tanks. Solutions of ferrocene dissolved in toluene at different concentrations are 159

poured into these tanks connected to an injection system and an evaporator where they are 160

nebulized and carried to the reactor by a gas flow of 1 L/min at a rate about 0.8 g/min. The gas 161

flow is a mixture of Ar and 30 vol.% of H2. A solution with 2.5wt.% of ferrocene is poured into

162

tank A (Fig. 1) while in tank B, the solution has variable concentrations. The solution in tank 163

A is injected for 1 min 40 s and then the solution in tank B for 13 min 20 s. The switch from 164

tank A to tank B is instantaneous, so that the process remains a one-step continuous process. In 165

single injection experiments, only the solution in tank B is injected for 15 min, with ferrocene 166

concentrations ranging from 0.01 to 2.5wt.%. 167

168

Fig. 1: Experimental set-up used for the one-step aerosol-assisted CCVD process operated

169

at atmospheric pressure. Insert: temperature profile at 800°C measured with a

170

thermocouple moved along the quartz reactor.

171

2.2 Characterizations

172

Morphology and thickness of VACNT carpets grown on the substrate and on the walls in the 173

center of the reactor are investigated by SEM (Carl Zeiss Ultra 55). CNT structure and statistical 174

analysis of external and internal diameters are obtained from TEM (Philips CM12, 120 kV) 175

after ultrasonic dispersion of CNTs in ethanol and deposition on standard TEM holder (copper 176

grids). For high resolution TEM, we use a JEOL 2011 operating at 200 kV, equipped with a 177

CCD camera (GATAN system ORIUS SC100, 4008×2672 pixels, pixel size: 0.014 nm). 178

TGA under air of products collected in the area in the center of the reactor is performed with a 179

92–16.18 SETARAM apparatus, at a temperature rate of 10°C/min in the 20-1000°C range, 180

which leads to the complete gasification of carbon. The rusty residue is used to determine the 181

iron content of the sample (weight percentage of iron in the collected products), as well as the 182

catalytic yield (ratio between the carbon and iron masses). This iron content is calculated 183

knowing that the residue consists of Fe2O3 as shown by X-ray diffraction [36]. In the unlikely

184

event that FeO contributes to the residue, the error would be small since the mass of iron in 185

Fe2O3 is 70%, while it is 78% in FeO. We will see in the following that this error would be

186

negligible given the magnitude of the variation in iron content in our experiments. 187

Raman spectra are acquired from the substrate samples in ambient conditions using a Renishaw 188

INVIA microspectrometer equipped with an Ar laser (514.5 nm wavelength), focused through 189

a Leica microscope. A very low incident power (2 mW) is used to avoid heating effects. Spectra 190

are recorded in the middle of the carpet cross-section in the 1000-3500 cm-1 _{range. Pertinent}

191

quantitative Raman parameters are obtained by conventional fitting procedures [37] for the 192

classical G, D, D’ and 2D bands [38] and the corresponding position, full-width at half 193

maximum (FWHM) and intensity (band height) relative to the G band are reported in Table S1

194

(Supplementary data). 195

3 Results

196

3.1 Thickness and growth rate of the VACNT forests

197

Double injection experiments

198

Double injection syntheses are performed with a first injection sequence with 2.5wt.% ferrocene 199

in toluene during 1 min 40 s immediately followed by a long injection sequence of 13 min 20 s 200

with a ferrocene concentration varying between 0.01 and 1.5wt.%. In preliminary experiments, 201

we checked that the 1 min 40 s duration is long enough to initiate a VACNT carpet 202

representative of a VACNT growth with a ferrocene concentration of 2.5wt.% (See Fig. S1 in 203

supplementary data). SEM images of the products grown on quartz substrates placed in the 204

oven center (Fig. 2a) reveal VACNT forests of the same good quality as usually reported [34]. 205

The shortest carpet (170 µm) is obtained for 0.01wt.% of ferrocene in the second injection 206

sequence, but then, the thickness increases sharply to 330 µm for 0.05wt.% and up to 380 µm 207

for 0.1wt.%, after which it remains around 270 µm for higher concentrations. The 208

corresponding growth rate (Fig. 2b) exceeds 25 µm/min at 0.1wt.% of ferrocene and stabilizes 209

between 15 and 20 µm/min at higher concentrations. 210

211

Fig. 2: Double injection. (a): SEM images of the VACNT carpets grown on the central

212

substrate (side-view) for increasing ferrocene concentration in the second long injection

213

sequence. (b): mean carpet thickness and growth rate on the substrate (red dot) and on

214

the reactor walls (blue triangle) as a function of ferrocene concentration. Dashed lines are

215

added as eyes guide only. The thickness distribution was measured by SEM on the same

216

samples with 20 measurement points for the CNTs on the walls of the reactor and 3

217

measurement points for the CNTs on the substrate.

218

The values on the substrate are slightly higher than those obtained on samples collected on the 219

reactor walls. (Fig. 2b). It is because the substrate is placed upstream along the reactor of the 220

zone where the samples used for the SEM analysis are collected and that, in aerosol-assisted 221

CCVD, the growth yield decreases along the reactor due to the dynamics of the catalytic 222

nanoparticles deposition as observed previously by us [15,34] and others [36]. This effect is 223

confirmed by the measurement of the mass distribution along the reactor displayed in Fig.S2. 224

As detailed in Supplementary Data, Fig.S2 shows that, whatever the ferrocene concentration, 225

the yield is optimal in zone 2 where the substrate is placed while the collected zone considered 226

in Fig.2 is zone 3 where the yield is slightly lower. 227

The most striking result of these double injection experiments is that, in the presence of H2, a

228

ferrocene concentration as low as 0.01wt.% in the second injection sequence is sufficient to 229

sustain the growth, and a rate higher than 15 µm/min is obtained for a concentration from 230

0.05wt.%. 231

Single injection experiments

232

Considering the very low ferrocene concentration (0.01 wt%) able to sustain the growth in 233

double injection synthesis, single injection synthesis with ferrocene concentration from 0.01 to 234

2.5wt.% are performed (Fig. 3a). For 0.01 and 0.05wt.% ferrocene, only a thin layer of 235

entangled nanotubes is observed. However, as soon as the ferrocene concentration reaches 236

0.1wt.%, the growth of a well-defined VACNT forest becomes effective with a growth rate as 237

high as 20 µm/min, close to its maximum of 22 µm/min obtained with 0.25wt.% (Fig. 3b). For 238

higher concentrations, the growth rate decreases to values in the 10-15 µm/min range. The 239

trends are the same on the substrate located at the center of the furnace and on the reactor walls, 240

with still mean values slightly lower for samples collected on the walls, due to variation along 241

the reactor (Fig. S2). The VACNT carpet thickness increases linearly with the synthesis 242

duration in the tested duration range up to 420 min, with a growth rate of 11.4 µm/min at 243

1.25wt.% and 2.5wt.% ferrocene (Fig. S3 in supplementary data). 244

245

Fig. 3: Single injection. (a): SEM images of the VACNT carpets grown on the central

246

substrate for increasing ferrocene concentration. (b): mean carpet thickness and growth

247

rate on the central substrate (red dot) and on the reactor walls (blue triangle) as a function

248

of ferrocene concentration. Dashed lines are added as eyes guide only. The thickness

249

distribution was measured by SEM on the same samples with 20 measurement points for

the CNTs on the walls of the reactor and 3 measurement points for the CNTs on the

251

substrate.

252

Therefore for this one-step aerosol assisted CCVD process operated at 800°C with hydrogen in 253

the gas phase, the ferrocene concentration threshold allowing an efficient growth of VACNTs 254

is as low as 0.1wt.%, while no growth at all was observed without hydrogen at 850°C for 255

ferrocene concentration equal to 0.1 or 0.5wt% [15]. More generally, the present results 256

represent a great improvement as compared to the state of the art related to the one-step CCVD 257

process (Table 1) in terms of growth rate at very low ferrocene concentration. 258

3.2 Purity, diameter and structure of carbon nanotubes

259

The main characteristics of the VACNTs are similar regardless of the injection mode and 260

depend only on the ferrocene concentration. Results obtained for the single injection mode will 261

be discussed here and the ones regarding the double injection reported in Supplementary data. 262

Let’s focus on single injection samples and compare SEM and TEM images of the formed 263

nanotubes with 0.1wt.% and 2.5wt.% of ferrocene respectively (Fig. 4). The striking difference 264

between both samples is in terms of cleanliness. At 0.1wt.%, the forest is very clean and no 265

parasitic particles can be observed outside the nanotubes at the resolutions used for the 266

observations. There are very few iron-based nanowires inside the nanotubes (Fig. 4a and 4c). 267

The number of embedded Fe nanoparticles has been drastically reduced as compared to our 268

previous results. Indeed, TEM micrographs on figure 4c) clearly indicate a high purity of CNT 269

at low ferrocene concentration. On the contrary, at 2.5wt.% numerous iron-based particles 270

exhibiting a nanowire shape are located inside the nanotubes (see also Fig. S4). In addition, a 271

considerable amount of nanoparticles are located on CNT surface (Fig. 4b and 4d), as observed 272

by others when the ferrocene concentration is high [19,21,33]. They are of varying sizes, some 273

of them being encapsulated into carbon shells and some others giving rise to the growth of 274

nanotubes of very small diameter which represent a low fraction of the sample mass and are not 275

considered in the diameter distributions reported below. Such iron nanoparticles located on the 276

CNT surface and the thin nanotubes observed on samples prepared with 2.5wt.% of ferrocene 277

(Fig. 4d) cannot be detected in the sample obtained with 0.1wt.% of ferrocene (Fig. 4c). 278

Therefore, lowering the ferrocene concentration induces a drastically reduction of the iron 279

impurities inside the VACNT samples. 280

A detailed HRTEM analysis of a few individual nanotubes over their length (see Fig. S4) is 281

used to characterize the iron-based nanowires found in the inner channel of the tubes. Their 282

average length and number density decrease significantly as the ferrocene concentration 283

decreases. This detailed TEM analysis at higher magnification (Fig S4) shows also a highly 284

reduced frequency of incorporation of nanowires into the tubes as well as a much lower mean 285

size of these nanoparticles when the ferrocene concentration is reduced. Therefore, the amount 286

of embedded nanoparticles is really low, and it is a great improvement for a one-step CCVD 287

process. For 0.1wt.% ferrocene, only 2% of the length of the nanotube inner channel is occupied 288

by iron nanowires, while it is 24% for 2.5wt.%. On average, it can be estimated that there is 1 289

nanowire every 2 µm at 0.1wt.% ferrocene while there is one nanowire every 0.4 µm at 2.5wt.% 290

(See Table S2). An analysis of structural defects (corking) in the inner channel also shows a 291

decrease in their occurrence for 0.1wt.% as compared to 2.5wt.%, but not in the same proportion 292

as nanowires (Table S2). 293

294

Fig. 4: Comparison in single injection between samples obtained with ferrocene

295

concentration of 0.1 and 2.5wt.%. a) SEM image of 0.1%; b) SEM image of 2.5%; c) TEM

296

image of 0.1%; and d) TEM image of 2.5%.

297

The iron content of VACNT forests is determined by TGA analysis under air (Fig. 5a) on 298

samples taken in the same central area of the reactor as the ones used for the SEM and TEM 299

analysis above. In single injection, the iron content varies greatly and linearly with the ferrocene 300

concentration, from 0.5wt.% at 0.1wt.% ferrocene up to 9wt.% at 2.5wt.% ferrocene. This value 301

of 0.5 wt% of iron in VACNT sample is a remarkably low value as compared to the literature 302

(see Table 1 and references herein) and is even lower than that of 0.56 recently published in a 303

paper related to a two-step floating CCVD process[31]. It confirmed the great increase in the 304

VACNT purity at low ferrocene concentration as observed by TEM above. In double injection, 305

iron content variations are less pronounced as compared to single injection (Fig. 5a) due to the 306

first injection sequence at 2.5wt.%. To verify that the variation in iron content with the ferrocene 307

concentration is not due to a similar variation in iron chemical yield, we estimate the proportion 308

of injected iron found in the final product. In single injection, the incorporated iron represents 309

40% of the injected iron at 0.1wt.% ferrocene and 50% at 2.5wt.%. Therefore, the incorporation 310

rate of iron remains similar regardless of the ferrocene concentration and cannot be responsible 311

for the high difference in iron content. 312

The catalytic yield (Fig. 5b) scales as the inverse of the ferrocene concentration for single 313

injection. A very high and maximum catalytic yield about 200 (200g of carbon produced by 314

iron gram) is obtained for the lowest concentration of ferrocene (0.1wt.% ) and then decreases 315

rapidly down to 10 when the ferrocene concentration increases. For double injection, the 316

catalytic yield is varying in a slighter range, between 16 and 35, for the whole ferrocene 317

concentration range. 318

319

Fig. 5: Ferrocene concentration effect on a) iron content and b) catalytic yield, for single

320

(full dot) and double (open dot) injection (in this last case, it is the ferrocene concentration

321

of the second long injection sequence). Dashed lines are added as eyes guide only.

Regarding the CNT diameter, the distribution does not change with the ferrocene concentration 323

(Fig. 6a), with external and internal average diameters around 35 nm and 9 nm respectively. 324

Similar results are found in the case of double injection (Fig.S5). In terms of structural quality, 325

the Raman spectra of the samples obtained with 0.1 and 2.5 w.t% ferrocene are compared in 326

Fig. 6b (see Fig. S6 for double injection): G, D and D’ bands at 1580 cm-1, 1350 cm-1 and 327

1615 cm-1 respectively, as well as the 2D band at 2700 cm-1 are occurring. In multi-walled 328

carbon nanotubes, the D band is associated with defects and disorder and is influenced by the 329

curvature of the walls, while the G band is inherent to in plane tangential stretching of carbon-330

carbon bonds. The ratio between the D and G band intensities (ID/IG) is commonly used as a

331

qualitative measure of the defect density in defective tubes: the higher this ratio, the lower the 332

crystalline level [38,39]. This ratio is dependent on the excitation wavelength. The analysis of 333

the Raman spectra recorded at 514 nm on the samples obtained in the current study gives rise 334

to a ID/IG ratio remaining similar whatever the injection mode and the ferrocene concentration.

335

This ratio is around 0.3 and the lowest value of 0.28 is obtained at 0.25wt.% ferrocene 336

concentration in single injection (Table S1). Such values are low for multi-walled carbon 337

nanotubes produced by CVD and are related to a low defect density as discussed by Meysami 338

et al. [21]. These values compare well with those reported in Table 1 obtained with similar 339

Raman wavelengths (514 or 532 nm) for samples formed in similar conditions in terms of 340

carbon precursors and synthesis temperatures (750-850°C) [15,19,21]. By contrast, they are 341

much lower than those reported in the range [0.56-0.79] by Kinoshita et al. [31] from synthesis 342

at 700°C with acetylene as carbon source. However, it has been shown that the G band in CNT 343

can be affected by defects [39], therefore as the 2D band does not represent a disorder-induced 344

mode, the relative intensity of the D and 2D bands (ID/I2D) is also used as a qualitative measure

345

of the defect density. Kinoshita el al. [31] found a ID/I2D ratio between 0.8 and 2. In the present

346

work, this ratio is around 0.3 for almost all the ferrocene concentrations, apart for 2.5wt.% of 347

ferrocene which gives rise to a ratio of 0.36. Such low ID/I2D ratio indicates a really good

348

structural quality of the CNT forest that is well preserved and even slightly improved at low 349

ferrocene concentration. This result corroborates the same trend observed by Cho et al. [20] 350

and McKee et al. [19] with a similar one-step aerosol-assisted process. 351

352

Fig. 6: Comparison in single injection between samples obtained with ferrocene

353

concentration of 2.5 (red) and 0.1wt.% (blue), respectively. a) Nanotube external diameter

354

distributions from the TEM analysis of more than 100 CNTs; b) Raman spectra acquired

355

at 514 nm at the center of forest cross-sections.

356

3.3 VACNT forest density

357

The mass per unit area of VACNTs collected on the reactor walls in the oven center (Fig. 7a)

358

increases with ferrocene concentration for both single and double injection experiments, 359

meaning that the chemical yield improves in both cases. In single injection, the areal mass is 360

close to 2mg/cm2 at 0.1wt.% and then increases gradually up to 4 mg/cm2. In double injection, 361

the areal mass is already equal to 3 mg/cm2 at 0.01wt.% ferrocene in the second injection and 362

then rises up to more than 6 mg/cm2, with values systematically higher than in single injection. 363

The gap between the two curves remains approximately constant around 2.5 mg/cm2 in excess 364

in double injection as compared to single injection. This difference can be attributed to the 365

effect of the first injection sequence with 2.5wt.% in the double injection mode. Reproducibility 366

has been tested for experiments performed with ferrocene concentration of 0.1, 0.25 and 2.5 367

wt.% in single injection and 0.1, 0.25 and 0.5 wt.% in double injection. The observed deviations 368

do not affect the general trends, particularly for single injection experiments (Fig. S6 in

369

supplementary information). 370

From measurements of areal masses (Fig. 7a) and forest thicknesses (Fig. 2b andFig. 3b), the 371

forest density can be determined and increases with the ferrocene concentration, especially in 372

single injection (Fig. 7b). Since nanotube diameters are similar whatever the ferrocene 373

concentration, this density increase must be related to an increase in N, the number density of 374

nanotubes (CNTs/cm2). N can be estimated knowing the mean external diameter of the 375

nanotubes and taking 1.8 g/cm3 _{for the intrinsic carbon nanotube density as estimated by helium}

picnometry. N is found about ~5 109 _CNT/cm2 _{at low ferrocene concentration, but reaches}

377

~1.7 1010 _CNT/cm2 _{at 1wt.%. This result suggests that the number density of catalytic}

378

nanoparticles increases as well. 379

380

Fig. 7 – Ferrocene concentration effect on a) mass per unit area of carbon collected on the

381

reactor walls in the center zone, and b) measured mass density, for single (full dot) and

382

double (open dot) injection (in this last case, it is the ferrocene concentration of the second

383

long injection sequence). Dashed lines are added as eyes guide only.

384

3.4 Oxidation resistance

385

The TGA mass-loss profiles are very dependent on the ferrocene concentration (Fig. 8a; see

386

also Fig. S8). In single injection, the profile is strongly shifted towards the high temperatures 387

when the ferrocene concentration decreases. The oxidation temperature (Tox), taken as the

388

temperature of the peak in the TGA graph derivatives (Fig. 8b and Fig. S8), decreases quasi-389

linearly from 685°C down to 550°C when ferrocene increases from 0.1 to 2.5wt.%. Since in 390

single injection the iron content of the sample was previously found to be a linear function of 391

the ferrocene concentration (Fig 5a), Tox appears to be a linear function of the iron content as

392

well (Fig. 8c and Fig. S9). In the case of double injection, the TGA profile exhibits a remarkable 393

shoulder on the high temperature side when the ferrocene concentration in the second injection 394

sequence is lower than 0.5wt.%, so that two Tox (labelled Double1 and Double2) can be

395

identified at low ferrocene concentrations (Fig. 8c and S9). The difference between both values 396

is really significant (nearly 100°C) which points to the existence of two distinct components in 397

the samples as discussed below. 398

399

Fig. 8: a) TGA graphs obtained at a temperature rate of 10°C/min and b) their derivatives

400

for single injection samples (blue dashed lines labelled from S_0.1 to S_2.5 according to

401

the ferrocene concentration) compared to the double injection with 0.01wt.% ferrocene

402

in the second injection sequence (orange solid line labelled D_0.01). c) Oxidation

403

temperature as a function of iron content. In double injection the label “Double2”

404

corresponds to the second peak at high temperature when observed in the dTG graph.

405

See also Fig. S8.

406

4 Discussion

407

The present results show that the one-step aerosol-assisted CCVD process can be operated at 408

very low ferrocene concentration (0.1 wt.%) using an Ar/H2 gas mixture as carrier gas. Using 409

these conditions and a synthesis temperature of 800°C, we demonstrate that this one-step 410

process enables the formation with a high catalytic yield (200) of VACNT forests exhibiting 411

low impurity level (0.5 wt.% of Fe). Such low ferrocene concentration does not involve any 412

reduction of CNT growth rate since it remains higher than 20 µm/min. Compared to the 413

literature (Table 1), it is, to our knowledge, the lowest concentration of ferrocene reported so 414

far for aerosol-assisted CCVD process leading to such growth rates and impressive catalytic 415

efficiency, thus involving the formation of high purity VACNT forests. The growth of a 416

VACNT forest becomes impossible only when the ferrocene concentration falls below 0.1wt.%. 417

When it increases up to 1wt.%, the growth rate decreases by a factor of 2, while the number 418

density N is multiplied by more than 3 going from 5 109 to 1.7 1010 cm-2. 419

Regarding the mechanisms at play, hydrogen plays a significant role in the process of ferrocene

420

decomposition. Indeed, our previous results obtained at 850°C with 2.5wt.% ferrocene [34] 421

have shown that including 30% of H2 in the Ar gas flow induces an increase of N from 5 109 to

422

2 1010 CNT.cm-2 as well as a decrease of the mean CNT external diameter. These results were 423

attributed to the hydrogen induced enhancement of the dissociation rate of ferrocene. As a 424

consequence, there is a fast rise of the iron partial pressure in the reactor and a high level of 425

supersaturation which affects the homogeneous nucleation process of the iron nanoparticles. A 426

much higher number of smaller nuclei forms, which leads to a higher number density of 427

catalytic nanoparticles deposited on the substrate and therefore to a higher density of growing 428

CNTs. Such very efficient effect of hydrogen on the catalytic particles density is responsible 429

for the remarkable decrease in the ferrocene concentration allowing to reach the threshold in 430

density of CNTs needed to form a self-supporting forest structure [40]. Interestingly, in the 431

present work at 0.1wt.% ferrocene in single injection mode, N is equal to 5 109CNT.cm-2, a 432

value similar to the one obtained without hydrogen at 2.5wt.% ferrocene with the same one-433

step aerosol assisted CCVD process [18]. This number allowing the vertical self-organization 434

of the interacting CNTs is comparable to the value of about 2 109 CNTs/cm2 reported by 435

Bedewy et al. [40]. This threshold is obviously not reached below 0.1wt.%, explaining the 436

absence of VACNT growth at such low concentration. This threshold was also not reached 437

without hydrogen for ferrocene concentration below 1.25wt.%, explaining why no VACNTs 438

were observed at 0.1 and 0.5wt.% by Castro et al. [15]. 439

In double injection synthesis, the first injection sequence with 2.5wt.% ferrocene determines 440

the density of catalytic nanoparticles. So N is higher from the beginning of the precursor 441

injection and its evolution is less pronounced with the ferrocene concentration increase in the 442

second injection sequence, where a ferrocene concentration as low as 0.05wt.% is sufficient to 443

sustain the growth. 444

The existence of a clear threshold in the ferrocene concentration allowing the self-organization 445

of VACNTs is responsible for the sharp increase in the growth rate with ferrocene concentration 446

observed for single injection experiments (Fig. 3b). The slow decrease observed after the 447

maximum (Fig. 3b), as also observed by others [32,33], is generally attributed to an increase in 448

VACNT density. Call et al. [41] show that ferrocene concentration dictates forest density but 449

that forest thickness is limited by density presumably due to diffusion-limited precursor supply. 450

However, in our case, the forest thickness is found proportional to the synthesis duration at least 451

up to 90 min even at 1.25 wt.% ferrocene where the forest density is above 1010 CNT/cm2 (Fig.

452

S3). Indeed, the flow rate is too high and the average inter-tube distance too large (between 60 453

and 150 nm) to encounter feedstock resistance [42]. Same conclusion was reported by Cho et 454

al. [20] leading them to invoke, as Bai et al. [33], a weaker activity of the catalytic nanoparticles

455

due to their size increasing with ferrocene concentration [8]. However, in our study, the CNT 456

diameter remains similar whatever the ferrocene concentration, which implies that the catalytic 457

particle size remains also the same [8]. In the present range of ferrocene concentration, this size 458

is mainly governed by temperature, H2 content and nature of the substrate, all these parameters

459

being constant here. On the other hand, when the number density of nanotubes increases, the 460

amount of precursors available per nanotube decreases, in principle in inverse proportion to the 461

density. This is not exactly what we are observing here since the mass per unit area that is the 462

product of the two is not constant, but increases. Such result could be rationalized considering 463

that at very low forest density, not all the carbon precursors impinging the substrate area around 464

a given CNT (intertube distance is 150 nm) are able to interact with the catalyst particle, but 465

their proportion increases when the intertube distance decreases. Such effect could explain the 466

observed variation of growth rate. 467

Regarding the iron content of the VACNT forest, it increases linearly from 0.48 to 8.75 wt.% 468

with the ferrocene concentration in single injection, and the atomic ratio Fe/C in the forest 469

appears to be 10 times the Fe/C ratio in the precursors (Fig. S10). It means that the iron 470

conversion is 10 times higher than the carbon conversion. It should be noted that the 471

contribution of the catalyst nanoparticles themselves to the iron content is very low as is out of 472

principle the case for any VACNT forest (see its estimation in Supplementary Data). The iron 473

incorporated in the material is predominantly found as nanowires encapsulated in the inner 474

channel as a consequence of the continuous refreshment of the catalyst nanoparticles from the 475

iron-based nanoparticles continuously nucleated in the gas phase [18] in the one-step aerosol-476

assisted CCVD process. When the ferrocene concentration is decreased, the percentage of the 477

CNT inner channel length occupied by the nanowires (Fig. S4 and Table S2) drastically 478

decreases: it is estimated to be 10 times lower for the optimal concentration of 0.1wt.% 479

ferrocene as compared to 2.5wt.%, which corresponds to an encapsulation frequency which is 480

3 times lower.. Since the CNT growth rate is two-fold higher at 0.1wt.% than at 2.5 wt.%, it 481

means that the acceleration in the refreshment at high ferrocene concentration does not allow 482

to increase the growth rate but only leads to an increase of the iron content. 483

Iron can be incorporated in the forest also as nanoparticles found outside the nanotubes. This 484

contribution increases a lot when the ferrocene concentration increases, as observed by SEM 485

and TEM (Fig. 4) and by others [21,33]. Since TEM observations show that the amount of 486

external nanoparticles is negligible in the sample obtained at 0.1wt.% ferrocene, then we can 487

estimate from the TEM analysis of the inner channel (Table S2) that the iron inside the tubes 488

represents about 2/3 of the total iron content for the sample formed with 2.5wt% ferrocene. 489

Therefore, external nanoparticles contain at most a third of the iron content. Their presence in 490

large quantities is due to the large increase, with the ferrocene concentration, of the number 491

density of iron nanoparticles in the precursor gas flow going towards the substrate through the 492

carpet. To avoid the formation of such large proportion of by-products in the presence of H2,

493

the ferrocene concentration must be kept lower than 1wt.%. 494

Regarding now the size and crystalline structure of the nanotubes, both TEM and Raman 495

analysis show that their structure and diameter depend only slightly on the concentration of 496

ferrocene in the range studied here. The ID/IG ratio of 0.3 indicates a good structural quality 497

with a low defect density for MWCNTs as observed previously [15,21]. That is why it is not 498

possible to interpret the large variation in the oxidation temperature with the ferrocene 499

concentration in terms of variation of the structural properties. Even considering the position of 500

the G-band, which was found to correlate with Tox in the literature [43], it is not possible to

501

invoke such structure evolution, since here, the G-band position is very stable (Table S1). 502

Alternatively, the catalytic activity of the iron-based compounds inside the forests during the 503

TGA was also considered in the literature since it is able to change the kinetic parameters of 504

the oxidation [19,21,23–25]. In the present case, this second interpretation is strongly supported 505

by the strong correlation found between Tox and iron content in single injection synthesis

506

(Fig.8c). Also, the highest oxidation temperature (680°C) obtained for the sample with the 507

lowest iron content is found very similar to the one of VACNT sample freed from iron after a 508

heat-treatment at 2000°C [29]. Moreover, in double injection, the catalytic activity of the iron-509

based compounds allows to understand directly the presence of two components in the TGA 510

graphs. When the ferrocene concentration is very different between the two injection sequences, 511

two layers are formed with very different iron contents and different oxidation temperatures. 512

The highest oxidation temperature is close to those obtained with single injection with a similar 513

low ferrocene concentration. The large variation in oxidation temperature under dry air can be 514

correlated to the iron content and the double peak observed for sample synthesized by double 515

injection underlines the localized catalytic role of iron compounds in the oxidation reaction as 516

mentioned in [25]. It could be used to perform selective oxidation of the carpet top in order to 517

suppress the layer where the CNTs are entangled and to open the tubes. Also, double injection 518

experiments offer the possibility to control independently forest density and growth rate while 519

limiting the iron content. The forest density is determined by the ferrocene concentration in the 520

first injection sequence while the growth rate depends mostly on the ferrocene concentration in 521

the second injection sequence. 522

5 Conclusion

523

Our results show that, in the aerosol-assisted CCVD process, the addition of H2 in the carrier

524

gas results in a significant increase in the number density of catalytic nanoparticles, due to its 525

strong influence on the decomposition kinetics of the ferrocene. As a result, the formation of a 526

self-organized VACNT forest occurs at a much lower ferrocene concentration than without 527

hydrogen and a pristine forest of high purity is formed with high catalytic yield at a performing 528

growth rate. At 800°C, with 30% H2 in Ar, VACNT forest is formed with only 0.1wt.%

529

ferrocene (Fe/C=0.007at%), a value 25 times lower than the usual one of 2.5% optimized 530

without hydrogen. The iron content falls down to 0.5wt.%, so that the catalytic yield is strongly 531

enhanced by a factor of 20, while the growth rate is improved by a factor of 2 and the crystalline 532

structure remains of high quality. A forest density of 5 109 CNTs/cm2 appears to be the 533

threshold for the self-organization of the nanotubes to occur. This number density increases 534

with the ferrocene concentration, due to the increase in the number density of catalytic 535

nanoparticles. The iron content of the forest is directly proportional to the ferrocene 536

concentration. When it decreases, the frequency of encapsulation of iron nanowires in the inner 537

channel of the tubes decreases significantly and the oxidation resistance of the samples is 538

greatly improved, with an increase in oxidation temperature equal to 100°C. Finally, double 539

injection experiments appear to be an attractive approach in order to optimize independently 540

the density and the growth rate of the forest, since the density is fixed by the initial ferrocene 541

concentration in the first short injection sequence while the growth rate can be controlled by 542

the concentration in the second injection sequence. 543

Acknowledgement

544

The authors would like to thank M. Paternostre (CEA Saclay), P.-E. Coulon (LSI, Ecole 545

Polytechnique Palaiseau) and J. N. Rouzaud (ENS Paris) for access to TEM, HRTEM and 546

Raman analysis facilities, respectively. 547

548

Appendix A. Supplementary data

549

Supplementary data associated with this article can be found, in the online version, at 550

551

References

552

[1] A.M. Marconnet, N. Yamamoto, M.A. Panzer, B.L. Wardle, K.E. Goodson, Thermal 553

Conduction in Aligned Carbon Nanotube- Polymer Nanocomposites with High Packing 554

Density, ACS Nano. 5 (2011) 4818–4825. doi:10.1021/nn200847u. 555

[2] S. Lagoutte, P.-H. Aubert, M. Pinault, F.O. Tran-Van, M. Mayne-L ’hermite, C. Chevrot, 556

Poly(3-methylthiophene)/Vertically Aligned Multi-walled Carbon Nanotubes: 557

Electrochemical Synthesis, Characterizations and Electrochemical Storage Properties in 558

Ionic Liquids, Electrochim. Acta. 130 (2014) 754–765. 559

doi:10.1016/j.electacta.2014.03.097. 560

[3] K. Evanoff, J. Khan, A.A. Balandin, A. Magasinski, W.J. Ready, T.F. Fuller, G. Yushin, 561

Towards ultrathick battery electrodes: Aligned carbon nanotube-enabled architecture, 562

Adv. Mater. 24 (2012) 533–537. doi:10.1002/adma.201103044. 563

[4] H. Cebeci, I.Y. Stein, B.L. Wardle, Effect of nanofiber proximity on the mechanical 564

behavior of high volume fraction aligned carbon nanotube arrays, Appl. Phys. Lett. 104 565

(2014). doi:10.1063/1.4862273. 566

[5] J.R. Raney, F. Fraternali, C. Daraio, Rate-independent dissipation and loading direction 567

effects in compressed carbon nanotube arrays, Nanotechnology. 24 (2013) 255707. 568

doi:10.1088/0957-4484/24/25/255707. 569

[6] M. Kumar, Y. Ando, Chemical Vapor Deposition of Carbon Nanotubes: A Review on 570

Growth Mechanism and Mass Production, J. Nanosci. Nanotechnol. 10 (2010) 3739– 571

3758. doi:10.1166/jnn.2010.2939. 572

[7] R. Guzmán de Villoria, S.L. Figueredo, A.J. Hart, S.A.I. Steiner, A.H. Slocum, B.L. 573

Wardle, High-yield growth of vertically aligned carbon nanotubes on a continuously 574

moving substrate, Nanotechnology. 20 (2009) 405611. doi:10.1088/0957-575

4484/20/40/405611. 576

[8] V. Jourdain, C. Bichara, Current understanding of the growth of carbon nanotubes in 577

catalytic chemical vapour deposition, Carbon N. Y. 58 (2013) 2–39. 578

doi:10.1016/j.carbon.2013.02.046. 579

[9] R. Andrews, D. Jacques, a. M. Rao, F. Derbyshire, D. Qian, X. Fan, E.C. Dickey, J. 580

Chen, Continuous production of aligned carbon nanotubes: a step closer to commercial 581

realization, Chem. Phys. Lett. 303 (1999) 467–474. doi:10.1016/S0009-2614(99)00282-582

1. 583

[10] M. Mayne, N. Grobert, M. Terrones, R. Kamalakaran, M. Rühle, H.W. Kroto, D.R.M. 584

Walton, Pyrolytic production of aligned carbon nanotubes from homogeneously 585

dispersed benzene-based aerosols, Chem. Phys. Lett. 338 (2001) 101–107. 586

doi:10.1016/S0009-2614(01)00278-0. 587

[11] X. Zhang, a Cao, B. Wei, Y. Li, J. Wei, C. Xu, D. Wu, Rapid growth of well-aligned 588

carbon nanotube arrays, Chem. Phys. Lett. 362 (2002) 285–290. doi:10.1016/S0009-589

2614(02)01025-4. 590

[12] R.A. Afre, T. Soga, T. Jimbo, M. Kumar, Y. Ando, M. Sharon, Growth of vertically 591

aligned carbon nanotubes on silicon and quartz substrate by spray pyrolysis of a natural 592

precursor: Turpentine oil, Chem. Phys. Lett. 414 (2005) 6–10. 593

doi:10.1016/j.cplett.2005.08.040. 594

[13] C. Singh, M.S.P. Shaffer, A.H. Windle, Production of controlled architectures of aligned 595

carbon nanotubes by an injection chemical vapour deposition method, Carbon N. Y. 41 596

(2003) 359–368. 597

[14] P. Boulanger, L. Belkadi, J. Descarpentries, D. Porterat, E. Hibert, A. Brouzes, M. Mille, 598

S. Patel, M. Pinault, C. Reynaud, M. Mayne-L’Hermite, J.M.M. Decamps, Towards 599

large scale aligned carbon nanotube composites: an industrial safe-by-design and 600

sustainable approach, J. Phys. Conf. Ser. 429 (2013) 012050. doi:10.1088/1742-601

6596/429/1/012050. 602

[15] C. Castro, M. Pinault, S. Coste-Leconte, D. Porterat, N. Bendiab, C. Reynaud, M. 603

Mayne-L’Hermite, Dynamics of catalyst particle formation and multi-walled carbon 604

nanotube growth in aerosol-assisted catalytic chemical vapor deposition, Carbon N. Y. 605

48 (2010) 3807–3816. doi:10.1016/j.carbon.2010.06.045. 606

[16] M. Pinault, M. Mayne-L’Hermite, C. Reynaud, V. Pichot, P. Launois, D. Ballutaud, M. 607

Mayne-L’Hermite, C. Reynaud, V. Pichot, P. Launois, D. Ballutaud, Growth of 608

multiwalled carbon nanotubes during the initial stages of aerosol-assisted CCVD, 609

Carbon N. Y. 43 (2005) 2968–2976. doi:10.1016/j.carbon.2005.06.011. 610

[17] R. Xiang, G. Luo, W. Qian, Q. Zhang, Y. Wang, F. Wei, Q. Li, A. Cao, Encapsulation, 611

compensation, and substitution of catalyst particles during continuous growth of carbon 612

nanotubes, Adv. Mater. 19 (2007) 2360–2363. doi:10.1002/adma.200602468. 613

[18] P. Landois, M. Pinault, S. Rouzière, D. Porterat, C. Mocuta, E. Elkaim, M. Mayne-614

L’Hermite, P. Launois, In situ time resolved wide angle X-ray diffraction study of 615

nanotube carpet growth: Nature of catalyst particles and progressive nanotube alignment, 616

Carbon N. Y. 87 (2015) 246–256. doi:10.1016/j.carbon.2015.01.046. 617

[19] G.S.B. McKee, C.P. Deck, K.S. Vecchio, Dimensional control of multi-walled carbon 618

nanotubes in floating-catalyst CVD synthesis, Carbon N. Y. 47 (2009) 2085–2094. 619

doi:10.1016/j.carbon.2009.03.060. 620

[20] J. Cho, A.R. Boccaccini, M.S.P. Shaffer, The influence of reagent stoichiometry on the 621

yield and aspect ratio of acid-oxidised injection CVD-grown multi-walled carbon 622

nanotubes, Carbon N. Y. 50 (2012) 3967–3976. doi:10.1016/j.carbon.2012.03.049. 623

[21] S.S. Meysami, A. a. Koós, F. Dillon, N. Grobert, Aerosol-assisted chemical vapour 624

deposition synthesis of multi-wall carbon nanotubes: II. An analytical study, Carbon N. 625

Y. 58 (2013) 151–158. doi:10.1016/j.carbon.2013.02.041. 626

[22] P.X. Hou, C. Liu, H.M. Cheng, Purification of carbon nanotubes, Carbon N. Y. 46 (2008) 627

2003–2025. doi:10.1016/j.carbon.2008.09.009. 628

[23] I.W. Chiang, B.E. Brinson, R.E. Smalley, J.L. Margrave, R.H. Hauge, Purification and 629

Characterization of Single-Wall Carbon Nanotubes, J. Phys. Chem. B. 105 (2001) 1157– 630

1161. doi:10.1021/jp003453z. 631

[24] A.-R. Leino, M. Mohl, J. Kukkola, P. Mäki-Arvela, T. Kokkonen, A. Shchukarev, K. 632

Kordas, Low-temperature catalytic oxidation of multi-walled carbon nanotubes, Carbon 633

N. Y. 57 (2013) 99–107. doi:10.1016/j.carbon.2013.01.040. 634

[25] C. Wang, S. Guo, X. Pan, W. Chen, X. Bao, Tailored cutting of carbon nanotubes and 635

controlled dispersion of metal nanoparticles inside their channels, J. Mater. Chem. 18 636

(2008) 5782. doi:10.1039/b811560e. 637

[26] E.R. Edwards, E.F. Antunes, E.C. Botelho, M.R. Baldan, E.J. Corat, Evaluation of 638

residual iron in carbon nanotubes purified by acid treatments, Appl. Surf. Sci. 258 (2011) 639

641–648. doi:10.1016/j.apsusc.2011.07.032. 640

[27] B.H.R. Suryanto, T. Fang, S. Cheong, R.D. Tilley, C. Zhao, From the inside-out: 641

Leached metal impurities in multiwall carbon nanotubes for purification or 642

electrocatalysis, J. Mater. Chem. A. 6 (2018) 4686–4694. doi:10.1039/c7ta11257b. 643

[28] X. Yang, M. Yang, H. Zhang, J. Zhao, X. Zhang, Q. Li, Electro-purification of carbon 644

nanotube networks without damaging the assembly structure and crystallinity, Appl. 645

Surf. Sci. 442 (2018) 232–238. doi:10.1016/j.apsusc.2018.02.169. 646

[29] M. Pinault, M. Mayne-L’Hermite, C. Reynaud, O. Beyssac, J.N. Rouzaud, C. Clinard, 647

Carbon nanotubes produced by aerosol pyrolysis: growth mechanisms and post-648

annealing effects, Diam. Relat. Mater. 13 (2004) 1266–1269. 649

doi:10.1016/j.diamond.2003.12.015. 650

[30] W. Huang, Y. Wang, G. Luo, F. Wei, 99.9% Purity Multi-Walled Carbon Nanotubes By 651

Vacuum High-Temperature Annealing, Carbon N. Y. 41 (2003) 2585–2590. 652

doi:10.1016/S0008-6223(03)00330-0. 653

[31] T. Kinoshita, M. Karita, T. Nakano, Y. Inoue, Two step floating catalyst chemical vapor 654

deposition including in situ fabrication of catalyst nanoparticles and carbon nanotube 655

forest growth with low impurity level, Carbon N. Y. 144 (2019) 152–160. 656

doi:10.1016/j.carbon.2018.12.019. 657

[32] M.I. Ionescu, Y. Zhang, R. Li, X. Sun, H. Abou-Rachid, L.S. Lussier, Hydrogen-free 658

spray pyrolysis chemical vapor deposition method for the carbon nanotube growth: 659

Parametric studies, Appl. Surf. Sci. 257 (2011) 6843–6849. 660

doi:10.1016/j.apsusc.2011.03.011. 661

[33] X. Bai, D. Li, Y. Wang, J. Liang, Effects of temperature and catalyst concentration on 662

the growth of aligned carbon nanotubes, Tsinghua Sci. Technol. 10 (2005) 729–735. 663

doi:10.1016/S1007-0214(05)70142-5. 664

[34] C. Castro, M. Pinault, D. Porterat, C. Reynaud, M. Mayne-L’Hermite, The role of 665

hydrogen in the aerosol-assisted chemical vapor deposition process in producing thin 666

and densely packed vertically aligned carbon nanotubes, Carbon N. Y. 61 (2013) 585– 667

594. doi:10.1016/j.carbon.2013.05.040. 668

[35] M. Pinault, V. Pichot, H. Khodja, P. Launois, C. Reynaud, M. Mayne-L’Hermite, 669

Evidence of sequential lift in growth of aligned multiwalled carbon nanotube 670

multilayers, Nano Lett. 5 (2005) 2394–2398. doi:10.1021/nl051472k. 671

[36] S.S. Meysami, F. Dillon, A. a. Koós, Z. Aslam, N. Grobert, Aerosol-assisted chemical 672

vapour deposition synthesis of multi-wall carbon nanotubes: I. Mapping the reactor, 673

Carbon N. Y. 58 (2013) 159–169. doi:10.1016/j.carbon.2013.02.044. 674

[37] E. Charon, J.-N. Rouzaud, J. Aléon, Graphitization at low temperatures (600-1200°C) in 675

the presence of iron implications in planetology, Carbon N. Y. 66 (2014) 178–190. 676

doi:10.1016/j.carbon.2013.08.056. 677

[38] M.A. Pimenta, G. Dresselhaus, M.S. Dresselhaus, L.G. Cancado, A. Jorio, R. Saito, 678

Studying disorder in graphite-based systems by Raman spectroscopy, Phys. Chem. 679

Chem. Phys. 9 (2007) 1276–1291. 680

[39] J. Maultzsch, S. Reich, C. Thomsen, S. Webster, R. Czerw, D.L. Carroll, S.M.C. Vieira, 681

P.R. Birkett, C.A. Rego, Raman characterization of boron-doped multiwalled carbon 682

nanotubes, Appl. Phys. Lett. 81 (2002) 2647–2649. doi:10.1063/1.1512330. 683

[40] M. Bedewy, E.R. Meshot, H. Guo, E.A. Verploegen, W. Lu, A.J. Hart, Collective 684

mechanism for the evolution and self-termination of vertically aligned carbon nanotube 685

growth, J. Phys. Chem. C. 113 (2009) 20576–20582. doi:10.1021/jp904152v. 686

[41] R.W. Call, C.G. Read, C. Mart, T.C. Shen, The density factor in the synthesis of carbon 687

nanotube forest by injection chemical vapor deposition, J. Appl. Phys. 112 (2012). 688

doi:10.1063/1.4768928. 689

[42] R. Xiang, Z. Yang, Q. Zhang, G. Luo, W. Qian, F. Wei, M. Kadowaki, E. Einarsson, S. 690

Maruyama, Growth deceleration of vertically aligned carbon nanotube arrays: Catalyst 691

deactivation or feedstock diffusion controlled?, J. Phys. Chem. C. 112 (2008) 4892– 692

4896. doi:10.1021/jp710730x. 693

[43] S. Santangelo, G. Messina, G. Faggio, M. Lanza, C. Milone, Evaluation of crystalline 694

perfection degree of multi-walled carbon nanotubes: Correlations between thermal 695

kinetic analysis and micro-Raman spectroscopy, J. Raman Spectrosc. 42 (2011) 593– 696

602. doi:10.1002/jrs.2766. 697